amazing graphics breakthrough

  • someone in chat linked this the other day

    it's really mindblowing:

    http://unlimiteddetailtechnology.com/description.html

    the short version is that instead of polygons it uses point clouds and the speed is only affected by the amount of pixels on screen, as no point is read or processed unless it will be shown on screen as a pixel:

    [quote:3kz5h6bu]Fig 1.4 shows some strange creatures with very complicated modeling, there are no straight or flat edges to be seen on them. Considering their high level models it is fair to say we could probably display about 4 of them on screen at once if you used today?s polygon system.(See Fig 1.5).....

    Here are thousands of them being displayed on an ordinary laptop, using only one core, No special 3D graphics hardware is being used, and it all runs in real-time.

    This is similar to id softwares id tech 5 engine, only the id engine is able to display textures of insane sizes with no affect on performance, unless the number of pixels on screen(resolution) changes :

    [quote:3kz5h6bu]

    using a more advanced MegaTexture approach called Virtual Texturing, which supports textures with resolutions up to 128000x128000 pixels

    both are amazing as hell. imagine them combined. nvidia said it's not impressed, which translates to, OH NO!!!

  • I saw a video about this technique some time ago.

    I'm not sure what to think. There is an extreme amount of skepticism about their claims (unsurprisingly) so until they have something that people can actually test for themselves it will probably just remain a curiosity.

    ...like ray tracing engines?

  • Until I see that "engine" running online with 20+ players all firing shots at each other, I don't trust it. They have no physics, characters or traditional effects (other than some flickering shadows and lighting).

    I prefer the Id Software approach. Taking the technologies we use today and making them more efficient.

  • Yeah, I'm not convinced either.

    It just seems too good to be true.

    We'll see.

    Krush.

  • Try Construct 3

    Develop games in your browser. Powerful, performant & highly capable.

    Try Now Construct 3 users don't see these ads
  • The page doesn't seem very professionally written, and with phrases like "I assure you, this works" rather than just demonstrating that it works with demos, it smells a bit fishy. Hallmarks of scam/self-promotion/I've-invented-a-perpetual-motion-machine.

    There might be a genuinely interesting algorithm behind this but it doesn't even appear to support lighting yet, so it looks a long, long way off replacing traditional engines. Also, I'm sure nVidia and ATI have been researching ideas like this like hell for years and probably figured out while cool, it can't replace everything existing technology does.

  • Well, I can't blame you guys for the skepticism. However...

    Subscribe to Construct videos now

    Also, apparently, according to news via Facebook, UD has been given actual funding and the go-ahead for development. It seems SOMEONE thinks its a good idea.

  • Well aside from the fact that the bird itself looked like garbage, I wasn't suggesting the engine couldn't do animation, just that they weren't showing it.

    They need to make a game, even if it looks like crap thanks to their "art."

    Also, I notice the system has the same problem voxels do, which is that things up close around the edges of the screen are extremely blocky.

    I just ... I don't like it. At all.

  • Hmm, I don't see nothing new here, and much less a breakthrough. That technique is well known, the points of a point cloud are called voxels (short for: volumetric pixel), and the first game engine working with that concept was voxel space by NovaLogic starting in the late 90's. There are a few games out there working with voxel space, the best known probably Delta Force.

    Today id software is thinking about doing a half voxel based engine following the very same principles for their next engine 6. Their trick is to just calculate static objects with the voxel technique while using conventional polygons for dynamic objects.

    In medical science voxels are a standard since the early 70's.

    Subscribe to Construct videos now

    Lighting is an issue, all voxel based engines need to use some form of baked lighting, although dynamic lighting adds so much to the gaming experience.

    I saw a few videos from unlimited detail. Every other phrase is "Well, we haven't done that yet, but we know it's possible." Of course they know, people have already done it.

    And

    Subscribe to Construct videos now

    for those who followed me to the end (

    )

  • Well, the breakthrough isn't the fact that its points instead of polys, its the fact that there isn't a limit to the detail. That hasn't ever been done with voxels. From what I understand voxels are slower than polys. If it is true, its a breakthrough. By your logic, th id tech 5 is nothing new because we've seen texture maps before

  • IWell, the breakthrough isn't the fact that its points instead of polys, its the fact that there isn't a limit to the detail. That hasn't ever been done with voxels. From what I understand voxels are slower than polys

    Of course that has been done before. That's what CT scans, x-rays etc. do every day

    The earlier voxel engines (like the first voxel space from 1997) were indeed slow. But that is more to the fact, that there weren't any hardware solutions. With the newer approach from id software a demo was presented running with stable 60 fps.

    The trick is to use something called a sparse voxel octree. Every voxel is assinged 8 subvoxels, every of them again 8 subvoxels etc. That way you limit the process of searching for points to be displayed within a huge point cloud.

    The real problem is the amout of raw data. Serious developers like id software don't ignore that. In fact, it is needed a technique to outsource that data to RAM and even to disc, because we are talking of an amount no PC in the next few years will have enough RAM for. That's not a problem if I only present a limited demo of a 20m forest without any characters (like unlimited detail did). But imagine a fps level with a few km, 20 detailed characters moving around, vehicles, etc...

  • Seems promising, I just hope they can get somewhere with it.

  • i hope this is not hoax because it will be a boon for people like me who have low end systems

  • Well, I can't blame you guys for the skepticism. However...

    Subscribe to Construct videos now

    I don't buy it still, a video could have been rendered over an hour. Better to see it running on your own machine at 30fps interactively to prove it. But still, as others have pointed out, it's plausible and has been done before and everything, it's just nowhere near as flexible as traditional 3D engines yet.

  • Until I see that "engine" running online with 20+ players all firing shots at each other, I don't trust it. They have no physics, characters or traditional effects (other than some flickering shadows and lighting).

    I prefer the Id Software approach. Taking the technologies we use today and making them more efficient.

    My brother saw this video and he said all graphic cards one day have to end one day to be replaced with this engine and whatever may come in the future, this engine could be the future of next-gen 3D gaming engine and the reason it has no physics and anything is because it is a prototype, this engine may bring a new meaning to 3D, now I don't know if I am correct on what I said above tough but from what I understood from it that is what I gathered, I do look forward for this and it is nowhere close to be used for a game right now.

  • I wish computer hardware wasn't so awful, then we wouldn't have to worry as much about these things.

Jump to:
Active Users
There are 1 visitors browsing this topic (0 users and 1 guests)